Aaron Li | Synth

01
Context Awareness

While applications such as ChatGPT and Claude are capable of answering general inquiries about facts and generating comprehensive answers, the responses are often not tailored to each individual user.

This means that the experience of conversing with an LLM falls apart when more personal, contextual questions are being asked.

A good example being:

“What was the central argument of the art history essay I wrote on Picasso last year?”

As a result, users often need to input large amounts of information for a desired output, often know as “prompting”.

01
Context Awareness

While applications such as ChatGPT and Claude are capable of answering general inquiries about facts and generating comprehensive answers, the responses are often not tailored to each individual user.

This means that the experience of conversing with an LLM falls apart when more personal, contextual questions are being asked.

A good example being:

As a result, users often need to input large amounts of information for a desired output, often know as “prompting”.

“What was the central argument of the art history essay I wrote on Picasso last year?”

01
Context Awareness

While applications such as ChatGPT and Claude are capable of answering general inquiries about facts and generating comprehensive answers, the responses are often not tailored to each individual user.

This means that the experience of conversing with an LLM falls apart when more personal, contextual questions are being asked.

A good example being:

As a result, users often need to input large amounts of information for a desired output, often know as “prompting”.

“What was the central argument of the art history essay I wrote on Picasso last year?”

Contextual Editing

As an editor, Synth can understand what you are doing at any moment, helping you become a better, more efficient writer.

Contextual Editing

As an editor, Synth can understand what you are doing at any moment, helping you become a better, more efficient writer.

Contextual Editing

As an editor, Synth can understand what you are doing at any moment, helping you become a better, more efficient writer.

02
Integration

During user interviews with new adopters of Large Language Models, many reflected the lack of continuity with existing workflows as a hindrance towards using AI applications more often.

Instead of an additional feature embedded within existing experience, new LLM applications are designed as standalone web interfaces, which creates a gap in the user flow.

This became a clear imperative to design Synth as an all-in-one integrated application, where language model AI would reside as a co-editor, and seamlessly function alongside other experiences.

02
Integration

During user interviews with new adopters of Large Language Models, many reflected the lack of continuity with existing workflows as a hindrance towards using AI applications more often.

Instead of an additional feature embedded within existing experience, new LLM applications are designed as standalone web interfaces, which creates a gap in the user flow.

This became a clear imperative to design Synth as an all-in-one integrated application, where language model AI would reside as a co-editor, and seamlessly function alongside other experiences.

02
Integration

During user interviews with new adopters of Large Language Models, many reflected the lack of continuity with existing workflows as a hindrance towards using AI applications more often.

Instead of an additional feature embedded within existing experience, new LLM applications are designed as standalone web interfaces, which creates a gap in the user flow.

This became a clear imperative to design Synth as an all-in-one integrated application, where language model AI would reside as a co-editor, and seamlessly function alongside other experiences.

Co-Editing

Synth is a first and foremost a co-editor. By integrating an LLM into your text editor, Synth can contextually understand text prompts and generate relevant outputs.

Co-Editing

Synth is a first and foremost a co-editor. By integrating an LLM into your text editor, Synth can contextually understand text prompts and generate relevant outputs.

Co-Editing

Synth is a first and foremost a co-editor. By integrating an LLM into your text editor, Synth can contextually understand text prompts and generate relevant outputs.

03
Text Heavy Interactions

When analyzing language model interfaces, we discovered that most interaction with LLMs rely heavily on textual inputs.

While this is beneficial for early adoption due to its flexibility in a wide range of use cases, text based interfaces often require large amounts of user input before reaching a desired outcome.

These additional interactions that could be eliminated with thoughtfully designed user experiences.

03
Text Heavy Interactions

When analyzing language model interfaces, we discovered that most interaction with LLMs rely heavily on textual inputs.

While this is beneficial for early adoption due to its flexibility in a wide range of use cases, text based interfaces often require large amounts of user input before reaching a desired outcome.

These additional interactions that could be eliminated with thoughtfully designed user experiences.

03
Text Heavy Interactions

When analyzing language model interfaces, we discovered that most interaction with LLMs rely heavily on textual inputs.

While this is beneficial for early adoption due to its flexibility in a wide range of use cases, text based interfaces often require large amounts of user input before reaching a desired outcome.

These additional interactions that could be eliminated with thoughtfully designed user experiences.

Visual Reasoning

Inspired by Maggie Appleton’s Language Model Sketchbook, one of the ways Synth breaks away from text heavy interfaces is it’s implementation of new interaction modalities with languages models. Instead of prompting, visual reasoning suggests possible logical relationships between ideas, and generates a tree like structure to visual concepts.

Visual Reasoning

Inspired by Maggie Appleton’s Language Model Sketchbook, one of the ways Synth breaks away from text heavy interfaces is it’s implementation of new interaction modalities with languages models. Instead of prompting, visual reasoning suggests possible logical relationships between ideas, and generates a tree like structure to visual concepts.

Visual Reasoning

Inspired by Maggie Appleton’s Language Model Sketchbook, one of the ways Synth breaks away from text heavy interfaces is it’s implementation of new interaction modalities with languages models. Instead of prompting, visual reasoning suggests possible logical relationships between ideas, and generates a tree like structure to visual concepts.

Takeaways

When analyzing language model interfaces, we discovered that most interaction with LLMs rely heavily on textual inputs.

While this is beneficial for early adoption due to its flexibility in a wide range of use cases, text based interfaces often require large amounts of user input before reaching a desired outcome.

These additional interactions that could be eliminated with thoughtfully designed user experiences.

Takeaways

When analyzing language model interfaces, we discovered that most interaction with LLMs rely heavily on textual inputs.

While this is beneficial for early adoption due to its flexibility in a wide range of use cases, text based interfaces often require large amounts of user input before reaching a desired outcome.

These additional interactions that could be eliminated with thoughtfully designed user experiences.

Takeaways

Reflecting on my journey with Synth, I realized the pressing need to transcend traditional text-based interactions. By integrating experiences such as visual reasoning and other non-textual modalities, we can create more intuitive and efficient interfaces that streamline interactions with LLMs. This project underscored the importance of developing novel, multimodal user experiences for the generative AI future.